De novo automated design of small RNA circuits for engineering synthetic riboregulation in living cells

Guillermo Rodrigoa,b,1, Thomas E. Landraina,1, and Alfonso Jaramilloa,2

aInstitute of Systems and , Genopole – Université d′Évry Val d′Essonne – Centre National de la Recherche Scientifique (CNRS UPS3509), 91030 Évry, France; and bInstituto de Biología Molecular y Celular de Plantas, Consejo Superior de Investigaciones Científicas – Universidad Politécnica de Valencia, 46022 Valencia, Spain

Edited by David Baker, University of Washington, Seattle, WA, and approved August 6, 2012 (received for review March 20, 2012)

A grand challenge in synthetic biology is to use our current knowl- tional problem, here we also rely on Watson–Crick interactions edge of RNA science to perform the automatic engineering of com- (i.e., canonical purine-pyrimidine pairs plus the G–U pair), pletely synthetic sequences encoding functional in living although it is possible to extend our approach to other types cells. We report here a fully automated design methodology and of interactions (28). Therefore, we are faced with the problem experimental validation of synthetic RNA interaction circuits work- of designing a set of RNA species with predefined structures ing in a cellular environment. The computational algorithm, based and with unspecified intermolecular interactions able to produce on a physicochemical model, produces novel RNA sequences by the intended allosteric regulation. Here, we show it is possible to exploring the space of possible sequences compatible with prede- solve such a combinatorial problem by developing a fully auto- fined structures. We tested our methodology in by mated procedure that exploits physicochemical principles and designing several positive riboregulators with diverse structures structural constraints and that outputs the RNA sequences imple- and interaction models, suggesting that only the energy of forma- menting the predefined interactions. tion and the activation energy (free energy barrier to overcome for initiating the hybridization reaction) are sufficient criteria to engi- Results and Discussion neer RNA interaction and regulation in bacteria. The designed se- Computational Design of sRNA Circuits. Our methodology (see de- quences exhibit nonsignificant similarity to any known noncoding tails in SI Appendix) starts by choosing well-defined structures for RNA sequence. Our riboregulatory devices work independently all single species of the circuit. Because we focus on Watson– and in combination with transcription regulation to create complex Crick interactions, we implicitly assume that the secondary struc- logic circuits. Our results demonstrate that a computational meth- ture already determines a stable three-dimensional architecture odology based on first-principles can be used to engineer interact- (29). We also hypothesize that the interaction between two spe- ing RNAs with allosteric behavior in living cells. cies is nucleated by their unpaired nucleotides (Fig. 1A). In this interaction model, an initial intermolecular pairing driven by a post-transcriptional regulation ∣ evolutionary computation ∣ small sequence (toehold or seed site) nucleates a downhill reac- computational RNA design ∣ RNA synthetic biology tion where the size of the intermolecular pairs (represented as a reaction coordinate in Fig. 1A) rapidly increases until the hybri- he understanding of RNA interactions in living cells and their dization ends. Initially, the algorithm assigns random nucleotides subsequent exploitation as regulators is providing new syn- to the sequences of each RNA species, while respecting their T ΔGi thetic biology applications (1). RNA regulation is being studied designated secondary structure, which ensures a low form by A from natural systems by the analysis of the interactions of small solving an inverse folding problem (Fig. 1 ). It then explores RNAs (sRNAs) with messenger RNAs (mRNAs) (2), the space of allowed nucleotide sequences by using an objective (3) or molecules (4). However, it is also possible to follow a for- function and a Monte Carlo simulated annealing (MCSA) search B ward engineering approach and attempt the de novo design of algorithm (30) (Fig. 1 ). The convergence of our algorithm is RNA regulators. Rational design techniques have been applied, coupled to the existence of large networks of neutral paths (of in both and , for repression or activation common structure) able to connect highly different sequences of (5–10), mRNA degradation (11), and (31), and the algorithm scatters several initial random sequences (12–14), transcription attenuation (15–18), and scaf- along these networks to perform an efficient exploration of the folding (19). On the other hand, computational methods allowed sequence space. In addition, to enlarge such neutral paths and designing nucleic-acid-based logic circuits in vitro (20–23), in- then improve the optimization, we allow non-neutral mutations cluding the redesign of allosteric ribozymes (23), hence, challen- perturbing, up to three base pairs, the structures specified for the ging the current knowledge of structure and function. single species. Previous RNA design approaches, however, have been mostly de- The objective function is defined to minimize by MCSA two i veloped to work with in vitro systems or required incorporating competing design goals: ( ) free energy of complex formation ii fragments of natural sequences. and ( ) activation energy of complex formation. The first term We now propose a fully automated sequence selection meth- accounts for the free energy difference between the interacting odology to design general circuits based on RNA interactions to and free species for all possible interactions in the circuit ΔG operate in living cells. Previous computational methodologies re- ( form). For the second term we consider a magnitude related lied on the dominance of the Watson–Crick interactions (24), but they were not adapted to in vivo operations where RNA could be Author contributions: G.R., T.E.L., and A.J. designed research, performed research, very unstable as it occurs in bacteria. Our approach consists of contributed new reagents/analytic tools, analyzed data, and wrote the paper. stabilizing RNA molecules by enforcing a given structure, as done The authors declare no conflict of interest. – in the inverse folding problem (25 27), together with targeted This article is a PNAS Direct Submission. interactions and conformational changes. The analysis of natural 1G.R. and T.E.L. contributed equally to this work. systems unveils another challenge: The kinetics of RNA interac- 2To whom correspondence should be addressed. E-mail: alfonso.jaramillo@issb tions is rate-limited by the initial interaction of solvent-accessible .genopole.fr. BIOPHYSICS AND nucleotides from each binding partner, as illustrated in the kis- This article contains supporting information online at www.pnas.org/lookup/suppl/

sing-loop mechanism (2). To make manageable the computa- doi:10.1073/pnas.1203831109/-/DCSupplemental. COMPUTATIONAL BIOLOGY

www.pnas.org/cgi/doi/10.1073/pnas.1203831109 PNAS ∣ September 18, 2012 ∣ vol. 109 ∣ no. 38 ∣ 15271–15276 Downloaded by guest on September 29, 2021 A Unfolded state Reaction scheme of modified models harnessing experimental data (33) can be em- RNA interaction RBS ployed to improve the prediction of the functionality of the ribor- Transition state egulatory devices. RBS G i form Engineering Synthetic Riboregulation. As a case study for our meth- i G act odology we chose the challenging problem of engineering a syn- Free Energy thetic riboregulator able to trans-activate the translation of a Intermolecular Individual Gform cis SI Appendix folding state -repressed gene ( , Fig. S4). This problem provides folding state RBS an assessment of the generality of our methodology because it requires the design of two RNA species that experience a confor- 0 dinteract Reaction Coordinate mational change after their mutual interaction. In this case, the quantification of the regulatory activity by differential ex- B Optimization Individual folding pression already results in a characterization of the RNA inter- scheme action. Furthermore, our riboregulators enlarge the repertoire of Scoring

RBS exogenous activators of bacterial for creating Sequences G = G + G interact act form gene circuits with complex dynamics (34). To design such ribor- sRNA Subjected to allosteric regulation Gconstr RBS Intermolecular folding << egulatory devices, we implemented a natural mechanism (35, 36), 5’-UTR mRNA 0 1 RBS = + ’ Gscore Ginteract (1 ) Gconstr where the sRNA activates translation by binding to the 5 -un- translated region (UTR) of a given mRNA. This binding is de- vised to produce a conformational change in the cis-repressed

Mutations binding site (RBS) to become exposed to the solvent and, therefore, enable the docking of the 16S ribosomal unit dur- FinP-like ing translation initiation (SI Appendix, Fig. S4). In our designs, we C T1 T2 T5 30 20 90 C1 SokC-like 40 80 ’ 40 60 maintained as fixed the RBS sequence in the 5 -UTR. We se- 30 T4 50 lected an mRNA coding for a fluorescent protein (GFP) to facil- 100 70

CU G C itate the experimental characterization. The secondary structures A A 20 10 AUU GUC A A U 30 60 U A 50 30 50 110 A 70 A U A U ’ 20 A C G for both sRNA and 5 -UTR of the mRNA were specified as U G GFP 40 A U G G C G U 120 A 50 A U C U A structural constraints (Fig. 1 ), although these specifications are G 40 1 79 C G 1 10 55 G U U A G U G A T3 G C G U not strict and some perturbations in the structures are tolerated. G 40 20 60 U A DsrA-like A U 130 40 A U A A U U G A G C G U For the sRNA, we picked several known structures to assay the A 705050 U A G C 10 U A A U U G 140 30 U G versatility of our methodology: the predicted structures of the 10 G C 80 A U U A AUG 30 G C 1 10 7071 C G U A bacterial sRNAs SokC, FinP, and DsrA (T1, T2, and T3, respec- 50 G U 20 G C 1 55 ACAUCAAACGUCCAGCGCCUG UUUUUUU 150 ’ 110 20156 tively) and two artificial structures (T4 and T5) (6, 8). For the 5 - 1 60 87 Structure imposed for UTR of the mRNAGFP, we chose a conformation (C1) shown to the 5’-UTR mRNA Naturally-occurring Artificially-generated Structures imposed for the sRNA produce the intended repression (6). Then, our algorithm ex- plored all possible sequences compatible with such specifications Fig. 1. Schemes of methodology and designs. (A) Thermodynamic scheme (31, 37), searching a stable hybridization of the two RNAs where of RNA interaction, showing the different free energies at play and the pro- gression of the reaction. We define the reaction coordinate as the size of the RBS remained unpaired. intermolecular pairs (d). (B) Optimization scheme followed to design the RNA devices. (C) Secondary structures specified for the single species to RAJ11 RAJ12 mRNA mRNA

obtain different RNA devices. Nucleotides shown were maintained fixed; RBS RBS RBS RBS mRNA mRNA RBS sequence yellow colored. Different devices were designed by imposing different structures for the riboregulator.

k toeholds to the on of the RNA interaction, which we assume is determined α ΔG by the length of the toehold sequence ( ), where act (activa- tion free energy) is proportional to C–α, C being a constant (21). RAJ21 RAJ22 mRNA

RBS RBS To facilitate convergence, we incorporate into the objective func- RBS mRNA mRNA mRNA tion an additional term accounting for a given structural con- RBS ΔG straint ( constr). These three terms are then weighted resulting in a scalar optimization problem (Fig. 1B). Recent experimental screening of about 500 mutants (between cis and trans) of the IS10 antisense RNA system shows that ΔG and ΔG are the form act RAJ23 RAJ31 two principal predictors of RNA interaction (18), which provides mRNA mRNA RBS RBS significant support to our objective function. Here, we approxi- mRNA RBS RBS mRNA mate the folding free energy of a given species (single or complex) from the minimum energy conformation (MEC), discarding the subdominant conformations of the corresponding thermody- pseudoknot namic ensemble. This is supported by the agreement between the MECs and the in vivo results recently reported for 3,000 yeast Fig. 2. Schematic representation of the six different RNA devices we de- transcripts (32). Our sequence selection procedure optimizes si- signed and engineered for riboregulation. Devices RAJ11 and RAJ12 were multaneously all RNA sequences of the circuit, and during the obtained by imposing the structure T4, device RAJ21 with T1, device optimization we do not need to impose natural nucleotide com- RAJ22 with T2, device RAJ23 with T3, and device RAJ31 with T5. SI – Appendix (Fig. S7) shows the helical structure of the different complexes positions or specific loop sequences (1 3), thus providing un- together with the corresponding base-pairing probability matrixes. SI biased synthetic sequences. In addition, because the evolutionary Appendix (Table S1) shows the sequences of species. SI Appendix procedure is independent of the RNA model, new energetically (Table S4) shows the thermodynamic properties of the systems.

15272 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1203831109 Rodrigo et al. Downloaded by guest on September 29, 2021 no sRNA sRNA instance, in system RAJ11 (SI Appendix, Fig. S6) a single point no RBS no cis-repression mutation would reduce the activity in 52% of the cases following our objective function (SI Appendix, Figs. S8 and S9). In particu- lar, we identified that two compensatory mutations in the toehold

Apparent would be sufficient to create an orthogonal device with similar activation fold specificity (SI Appendix, Table S8). 105 105 pSTC0 pSTC1 Context Analysis of Engineered Riboregulation. Our design metho- dology does not consider cellular factors that could drive RNA 104 104 interactions, such as the specific and nonspecific binding to en- dogenous RNAs or proteins. On the one hand, we could incor- porate into the designed sequences specific recognition sites to known proteins, such as Hfq, provided its role in sRNA stabiliza- 103 103 tion and catalysis of sRNA–mRNA pairing (38, 39). Instead, by expressing both RNAs within the same plasmid we may already

Fluorescence (normalized) promote local coexpression, avoiding intracellular RNA diffu- 102 102 sion. Other cellular factors involve ribonucleases (RNases), which could be incorporated if their cleavage sequences are known. On the other hand, nonspecific interactions could jeopar- Fig. 3. Experimental characterization of the RNA devices. Normalized dize our predictions. For instance, bacterial RNase III is a potent fluorescence of the devices together with the apparent activation fold (Top), and fast RNase that targets double stranded regions and we eval- measured as the ratio of fluorescence in presence and absence of the uated the effect of such an RNase on this system by replicating riboregulator (see also SI Appendix,Fig.S15, for flow cytometry character- izations). our characterizations in the corresponding knockout strain. Although we found an increase about the double in the apparent Δ We experimentally validated in E. coli six solutions corre- activation fold (from 11.2 to 26.7) when using a RNase III strain A sponding to different interaction models (Fig. 2) by exploring (Fig. 4 ), the real-time quantitative PCR (RT-qPCR) quantifica- ∕ 1040 SI Appendix cis tion of the ratio sRNA mRNAGFP did not change in such a strain a space of sequences ( , Table S7). The -repres- SI Appendix sing elements tightly reduced the protein expression to about 1– and it warrants further exploration (see ). 4% of the maximal expression (SI Appendix, Fig. S13). These Because our methodology finds novel nucleotide sequences stabilizing a set of RNA structures and interactions, we would strong repressions certainly support the formation of the pre- expect each sequence was sufficiently dissimilar to the others dicted structures, which prevent the exposure of the RBS to making cross-talk unlikely. Moreover, a Basic Local Alignment the solvent. The designed riboregulators activated translation Search Tool showed that all sequences of our devices display (monitored by fluorometry) in a range going from 2.8 (RAJ21) A no significant similarity to any known noncoding RNA sequence to 11.2 (RAJ11) of apparent activation fold (Fig. 3 ). These va- (SI Appendix, Table S6). To investigate the orthogonality of our lues are similar to those reported for natural systems (35); there- riboregulatory devices, we checked the hybridization ability be- fore, our devices could be readily used to reprogram bacterial tween possible combinations of cis-repressing and trans-activating behavior. A flow cytometry analysis also revealed a statistically RNAs. We estimated computationally the relative levels of the significant sRNA activation within the cell population in all de- interaction complex at the equilibrium, showing that our devices, signs (SI Appendix, Fig. S15), with average values presenting despite the homologies already present in the sequences and quantitative agreement with fluorometry results (SI Appendix, structures due to imposing a common RBS sequence, displayed Fig. S12). The significant activation of protein expression in all low interactions between noncognate pairs (Fig. 4B; SI Appendix, cases supports the sRNA-mRNA interaction and the intended Table S5; degree of orthogonality of the 98%). We also investi- ’ conformational change in the 5 -UTR of the mRNAGFP. Further- gated the effect of higher RNA concentrations on the orthogon- more, manual design can be applied on top of our computational ality, showing a notable dose-dependent cross-talk between designs to enlarge the number of different devices (16). For devices RAJ12 and RAJ21 (SI Appendix, Fig. S11). We then

A C no sRNA sRNA no RBS no cis-repression Apparent Apparent activation fold activation fold 5’-UTR 104 105 B cisRAJ21 cisRAJ23 cisRAJ31 cisRAJ11 cisRAJ12 cisRAJ22 100 104 transRAJ11 transRAJ12 103 transRAJ21 50 3

10 sRNA transRAJ22

transRAJ23 Fraction bound Fluorescence (normalized)

Fluorescence (normalized) transRAJ31 102 102 0

JS006 HT115

Fig. 4. Context analysis of the RNA devices. (A) Comparison of the activity of the device RAJ11 in the regular strain (JS006) and in a strain ΔRNase III (HT115). (B) Orthogonality analysis of the devices, showing the fraction of complex formation at the equilibrium (values provided in SI Appendix, Table S5; see also BIOPHYSICS AND

SI Appendix, Fig. S11). (C) Experimental validation of the orthogonality between the devices RAJ11 and RAJ12. COMPUTATIONAL BIOLOGY

Rodrigo et al. PNAS ∣ September 18, 2012 ∣ vol. 109 ∣ no. 38 ∣ 15273 Downloaded by guest on September 29, 2021 A AND logic gate

TetR aTc transRNA RBS PLtetO-1 RBS IPTG LacI 5’-UTR 5’-UTR GFP

PLlacO-1

IPTG=1 mM & aTc=100 ng/mL B x 104 C 2 IPTG=0 & aTc=100 ng/mL IPTG=1 mM & aTc=0 IPTG=0 & aTc=0 1.5

1

0.5 Fluorescence (normalized) 0 0 10 20 30 time (h)

Fig. 5. The RNA device can work in combination with transcription regulation. (A) Scheme of a circuit coupling these two types of control (resulting in an AND logic gate), where IPTG and aTc are the two inputs and GFP the output. To implement this circuit we used the device RAJ11. (B) Experimental characterization of the circuit showing the dynamics of the normalized fluorescence after introduction of inducers at time 0 (see also SI Appendix, Fig. S19, for the full set of dynamics at different levels of IPTG and aTc). (C) Arbitrary units of GFP (colored circles) are estimated by taking the maximum value of the dynamics shown in SI Appendix (Fig. S19), ensuring OD600 < 0.4, and dividing this value by the one obtained in case of no inducers in the medium. Solid lines are obtained with the mathematical model presented in SI Appendix (Eq. S7), with parameters fitted against these experimental data.

performed an experimental validation of these orthogonality Conclusions inferences with the devices RAJ11 and RAJ12. There we found In summary, we have presented a general methodology based that, despite having similar specificity, performance, and struc- on theoretical principles and combinatorial optimization to design ture, the noncognate interaction (transRAJ11 and cisRAJ12) dis- interacting RNAs. The ability of our simple model to guide the played no significant activity (Fig. 4C). We concluded that it was design of fully synthetic riboregulation suggests that intracellular the large size of the sequence space (SI Appendix, Table S7) that RNA interactions are predominantly governed by the energy of formation and the activation energy. Our de novo design approach permitted highly specific sRNAs to be obtained, which would al- relies on the enforcement of given structures to each RNA species, low them to operate within a same cellular compartment. which provides the required stability in the cytoplasm and guides allosteric regulation. However, our objective function promotes Engineering Modular AND Logic Gate. Our devices can be then com- the desired interactions, and the allosteric behavior is enforced bined with transcription regulatory elements to engineer combi- through a constraints term. We have validated the methodology natorial logic gates (6, 7). Because transcription regulation is also by engineering in bacteria different interaction models implement- very specific, we could create a large library of orthogonal logic ing translation activators. These tests already have addressed im- gates. Toexemplify this we designed an AND logic gate by placing portant challenges in the design of complex RNA interaction 5 0 the sRNA and the -UTR-mRNAGFP under the control of tun- circuits, such as the design of precise conformational changes A P P able promoters (Fig. 5 ). We used the LtetO-1 and LlacO-1 pro- and the interface with the cellular machinery. Although further moters together with a strain that constitutively expressed the tests of the methodology would strengthen the performance of transcription repressors TetR and LacI (40). Therefore, we could the algorithm, its basis on physicochemical principles would allow regulate the transcription of these promoters using the inducers us to apply it in many different frameworks and obtain more anhydrotetracycline (aTc) and isopropyl β-D-1-thiogalactopyra- sophisticated systems. For instance, we designed a pair of ribore- SI noside (IPTG), respectively. To implement such a system, we con- gulators able to synergistically activate protein expression ( Appendix sidered the device RAJ11. The resulting AND logic gate had a ,Fig.S21), illustrating the versatility of our approach very low leakage and displayed a sigmoidal response to both and the ability to explore even larger spaces. As our automated methodology uses few specifications as aTc and IPTG (Fig. 5 B and C and SI Appendix, Fig. S20). We inputs, it could also be used to test new mechanisms and hypoth- obtained the transfer function of the device by using the apparent eses despite the lack of a complete molecular understanding of A activation fold calculated from Fig. 3 . By setting high levels of the living cell. It would then be possible to average out the effects IPTG, the concentration of aTc allows tuning the activity of the of unknown natural systems by designing a large number of RNA device (SI Appendix, Fig. S19). Because riboregulators systems implementing a given set of specifications. On the other could be used with arbitrary promoters (6), we could create hand, the engineered RNA devices can be exploited in biotech- AND logic gates adapted to different applications. nology, in particular for metabolic control purposes, where the

15274 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1203831109 Rodrigo et al. Downloaded by guest on September 29, 2021 enzymatic expression is cis-repressed (10) and the riboregulator is measurements. Cultures were grown over-night at 37 °C and at 225 rpm from even combined with that sense a given metabolite (17, single-colony isolates before being diluted. The following concentrations of 41). As a result, the synthetic circuit can trigger the expression of antibiotics were used when appropriate: kanamycin (50 μg∕mL), tetracycline enzymes for regulating the activity of the pathway. Also in bac- (15 μg∕mL), spectinomycin (100 μg∕mL). When using E. coli K-12 MG1655-Z1 P teria, we could exploit other mechanisms such as ribozymes (14) cells, 1 mM of IPTG was used for full activation of the LlacO-1 promoter when 100 ∕ P or antiterminators (15), and high-throughput analyses could be needed, and ng mL of aTc was used for full activation of the LtetO-1 pro- moter when needed. addressed using a library of automatically designed sRNAs. Our methodology does not require any ribosome involvement, Fluorescence Quantification Using Fluorometry. Before characterization ex- where the RBS sequence could be replaced by the corresponding periments, cells were grown in M9 over two nights in order to reach station- binding sequence motif of an RNA-binding protein (or molecule) ary phase. Cultures were then diluted 200 times in 200 μL of M9 within each if known (19, 42). In addition, we could improve our methodology well of the plate (Custom Corning Costar 96-well microplate, black transpar- (see SI Appendix) by incorporating cellular factors (e.g., Hfq, ent bottom with lid). The plate was incubated in an Infinite F500 multiwell RNase III, RNase E, or 16S rRNA) into the physicochemical fluorometer (TECAN) at 37 °C with shaking (orbital mode, frequency of model, by considering non-Watson–Crick interactions (28), or 33 rpm, 2 mm of amplitude) and assayed with an automatically repeating by adding predictors of translation initiation by 5’-UTR se- protocol of absorbance measurements (600 nm absorbance filter) and fluor- quences (43). Furthermore, we could design riboregulators for escence measurements (480∕20 nm excitation filter–530∕25 nm emission fil- eukaryotic hosts by replacing the RBS by the Kozak sequence, ter for GFP and 580∕20 nm excitation filter–610∕10 nm emission filter for or alternatively, if known, by the internal ribosome entry site RFP). Time between repeated measurements was 15 min. All samples were – (44). In yeast, the sRNA would interact with the 5’-UTR of cer- present in 3 6 replicated on the plate. Each measurement was repeated tain mRNA to alter the cap-dependent scanning of the 40S sub- two to three times on independent days to verify reproducibility in the re- sults. All data analyses were done using values harvested when cells were in unit (45). In higher eukaryotes, the sRNA could target a given exponential growth phase (OD600 between 0.1 and 0.4). Growth rates were RNA, leading to a conformational change that triggers the in- obtained as the slope of a linear regression between the values of logðOD600Þ tended function (9). Yet, the proposed automatic algorithm to and time. Similar growth rates were observed in all experiments, except for design sRNA circuits in living cells will open new venues for the HT115 strain growing naturally slower. The steady-state protein expres- RNA synthetic biology (1) and for quantitative testing of our as- sion value was obtained as the slope of a linear regression between the sumptions about RNA function. values of fluorescence and OD600. For the dynamical analysis of protein ex- pression, the absolute fluorescence was divided by OD600 to have a magni- Materials and Methods tude per cell. Plasmid Construction. The SI Appendix (Figs. S1 and S2) outlines the two plas- mid templates, pSTC0 and pSTC1, used in this work. The SI Appendix Fluorescence Quantification Using Flow Cytometry. All expression data were (Table S11) shows all plasmids constructed. The target mRNA was placed collected using a Becton Dickinson FACSCanto II flow cytometer with a ’ P in 5 sense under the control of the LlacO-1 promoter (regulated by LacI 488 nm argon laser and a 530∕30 nm emission filter (GFP) and a 695∕40 nm ’ P and IPTG) and the sRNA in 3 sense under the LtetO-1 promoter (regulated emission filter (RFP). Overnight cultures in M9 were diluted 200 times in by TetR and aTc) (40). The RNA devices (from the terminator of the sRNA 200 μL of fresh medium and incubated to reach an OD600 of about 0.1. Cells ’ to the 5 -UTR of the mRNA) were made by DNA synthesis (DNA 2.0) and then from M9 cultures were fixed using 4% paraformaldehyde (PFA)—culture cells inserted by ligation into the plasmid templates. From the resulting plasmids, were pelleted and washed in filtered PBS (Biorad), washed 15 min with PFA, the sRNA was removed to generate systems bearing only the mRNA operon. washed in filtered PBS, and finally kept at 4 °C until measurement. Fluores- pSTC0 had the pMB1 origin, kanamycin resistance, and GFPmut3b as a repor- cence measurement of gene expression from each sample was obtained from ter gene. pSTC1 had the pSC101 origin, kanamycin resistance, and super- >20;000 cells. We analyzed the data using FCS express 4 (Denovo software), folder GFP as a reporter gene (see details in SI Appendix). and we gated the events using narrow forward and side scatter range. We then represented the fluorescence distributions in log scale. Strains, Reagents, and Cell Culture. E. coli TOP10 (Invitrogen) was used for rou- tine transformation, as described in the protocol (46). Characterization ex- periments were performed in E. coli K-12 JS006 cells (MG1655 ΔaraC ΔlacI ACKNOWLEDGMENTS. We thank J. Hasty for providing the E. coli strain JS006, KanS) (47), in E. coli K-12 HT115 cells (W3110 rnc-14::Tn10) as RNase III de- N. Guillén for the strain HT115, and M.B. Elowitz for the strain MG1655-Z1. We thank F. de la Cruz, R. Ruiz, and I. del Campo for practical support with pleted environment (48), and in E. coli K-12 MG1655-Z1 cells (MG1655 þ þ þ R flow cytometry measurements and S.F. Elena and M.P. Zwart for assistance in lacI tetR araC Sp ) for constitutive control over the P 1 and P 1 LlacO- LtetO- RT-qPCR assays. We also thank H. Isambert, J.J. Collins, F.J. Isaacs, B.F. Luisi, and – promoters (49). Cells were grown aerobically in Luria Bertani broth or in a E. Westhof for useful comments. Work was supported by the FP7-ICT-043338 modified M9 minimum media comprising M9 minimum salts (Sigma M6030) (Bacterial Computing with Engineered Populations), the Actions Théma- ∕ supplemented with glycerol at 0.8% (vol vol) as the only carbon source, tiques Incitatives de Genopole, and the Fondation pour la Recherche Medi- CaCl2 at 100 μM, MgSO4 at 2 mM, and FeSO4 at 100 μM for higher growth cale grants (to A.J.). G.R. is supported by an European yield (50). Casamino acids were avoided because of their natural green fluor- Organization long-term fellowship cofunded by Marie Curie actions (ALTF- escence that produces too high a fluorescence background for fluorometer 1177-2011), and T.E.L. by a PhD fellowship from the AXA Research Fund.

1. Isaacs FJ, Dwyer DJ, Collins JJ (2006) RNA synthetic biology. Nat Biotechnol 24:545–554. 11. Rinaudo K, et al. (2007) A universal RNAi-based logic evaluator that operates in mam- 2. Brantl S (2007) Regulatory mechanisms employed by cis-encoded antisense RNAs. Curr malian cells. Nat Biotechnol 25:795–801. Opin Microbiol 10:102–109. 12. Beisel CL, Bayer TS, Hoff KG, Smolke CD (2008) Model-guided design of ligand- 3. Rajkowitsch L, et al. (2007) RNA chaperones, RNA annealers and RNA helicases. RNA regulated RNAi for programmable control of gene expression. Mol Syst Biol 4:224. Biol 4:118–130. 13. Culler SJ, Hoff KG, Smolke CD (2010) Reprogramming cellular behavior with RNA 4. Gallivan JP (2007) Toward reprogramming bacteria with small molecules and RNA. controllers responsive to endogenous proteins. Science 330:1251–1255. Curr Opin Chem Biol 11:612–619. 14. Carothers JM, Goler JA, Juminaga D, Keasling JD (2011) Model-driven engineering of 5. Pfleger BF, et al. (2006) Combinatorial engineering of intergenic regions in operons RNA devices to quantitatively program gene expression. Science 334:1716–1719. tunes expression of multiple genes. Nat Biotechnol 24:1027–1032. 15. Dawid A, Cayrol B, Isambert H (2009) RNA synthetic biology inspired from bacteria: 6. Isaacs FJ, et al. (2004) Engineered riboregulators enable post-transcriptional control of Construction of transcription attenuators under antisense regulation. Phys Biol gene expression. Nat Biotechnol 22:841–847. 6:025007. 7. Callura JM, Dwyer DJ, Isaacs FJ, Cantor CR, Collins JJ (2010) Tracking, tuning, and 16. Lucks JB, Qi L, Mutalik VK, Wang D, Arkin AP (2011) Versatile RNA-sensing transcrip- terminating microbial physiology using synthetic riboregulators. Proc Natl Acad Sci tional regulators for engineering genetic networks. Proc Natl Acad Sci USA USA 107:15898–15903. 108:8617–8622. 8. Nakashima N, Tamura T (2009) Conditional gene silencing of multiple genes with anti- 17. Qi L, Lucks JB, Liu CC, Mutalik VK, Arkin AP (2012) Engineering naturally occurring sense RNAs and generation of a mutator strain of Escherichia coli. Nucleic Acids Res trans-acting non-coding RNAs to sense molecular signals. Nucleic Acids Res 37:e103. 40:5775–5786. 9. Venkataraman S, Dirks RM, Ueda CT, Pierce NA (2010) Selective cell death mediated 18. Mutalik VK, Qi L, Guimaraes JC, Lucks JB, Arkin AP (2012) Rationally designed families by small conditional RNAs. Proc Natl Acad Sci USA 107:16777–16782. of orthogonal RNA regulators of translation. Nat Chem Biol 8:447–454. BIOPHYSICS AND 10. Callura JM, Cantor CR, Collins JJ (2012) Genetic switchboard for synthetic biology 19. Delebecque CJ, Lindner AB, Silver PA, Aldaye FA (2011) Organization of intracellular

applications. Proc Natl Acad Sci USA 109:5850–5855. reactions with rationally designed RNA assemblies. Science 333:470–474. COMPUTATIONAL BIOLOGY

Rodrigo et al. PNAS ∣ September 18, 2012 ∣ vol. 109 ∣ no. 38 ∣ 15275 Downloaded by guest on September 29, 2021 20. Seelig G, Soloveichik D, Zhang DY, Winfree E (2006) Enzyme-free nucleic acid logic 36. Soper T, Mandin P, Majdalani N, Gottesman S, Woodson SA (2010) Positive regulation circuits. Science 314:1585–1588. by small RNAs and the role of Hfq. Proc Natl Acad Sci USA 107:9602–9607. 21. Yin P, Choi HM, Calvert CR, Pierce NA (2008) Programming biomolecular self-assembly 37. Sim AYL, Levitt M (2011) Clustering to identify RNA conformations constrained by – pathways. Nature 451:318 322. secondary structure. Proc Natl Acad Sci USA 108:3590–3595. 22. Ran T, Kaplan S, Shapiro E (2009) Molecular implementation of simple logic programs. 38. Papenfort K, Bouvier M, Mika F, Sharma CM, Vogel J (2010) Evidence for an autono- Nat Nanotechnol 4:642–648. mous 5’ target recognition domain in an Hfq-associated small RNA. Proc Natl Acad Sci 23. Penchovsky R, Breaker RR (2005) Computational design and experimental validation USA 107:20435–20440. of oligonucleotide-sensing allosteric ribozymes. Nat Biotechnol 23:1424–1433. – 24. Ellington AD (2007) What’s so great about RNA? ACS Chem Biol 2:445–448. 39. Vogel J, Luisi BF (2011) Hfq and its constellation of RNA. Nat Rev Microbiol 9:578 589. 25. Dahiyat BI, Mayo SL (1997) De novo protein design: Fully automated sequence selec- 40. Lutz R, Bujard H (1997) Independent and tight regulation of transcriptional units in tion. Science 278:82–87. Escherichia coli via the LacR/O, the TetR/O and AraC/I1-I2 regulatory elements. Nucleic 26. Jaramillo A, Wernisch L, Héry S, Wodak SJ (2002) Folding free energy function selects Acids Res 25:1203–1210. native-like protein sequences in the core but not on the surface. Proc Natl Acad Sci 41. Roth A, Breaker RR (2009) The structural and functional diversity of metabolite- USA 99:13554–13559. binding riboswitches. Annu Rev Biochem 78:305–334. 27. Koehl P, Levitt M (2002) Protein topology and stability define the space of allowed 42. Paige JS, Wu KY, Jaffrey SR (2011) RNA mimics of green fluorescent protein. Science – sequences. Proc Natl Acad Sci USA 99:1280 1285. 333:642–646. 28. Leontis NB, Stombaugh J, Westhof E (2002) The non-Watson–Crick base pairs and their 43. Salis HM, Mirsky EA, Voigt CA (2009) Automated design of synthetic ribosome binding associated isostericity matrices. Nucleic Acids Res 30:3497–3531. sites to control protein expression. Nat Biotechnol 27:946–950. 29. Flamm C, Fontana W, Hofacker IL, Schuster P (2000) RNA folding at elementary step resolution. RNA 6:325–338. 44. Hellen CU, Sarnow P (2001) Internal ribosome entry sites in eukaryotic mRNA mole- – 30. Kirkpatrick S, Gelatt CD, Vecchi MP (1983) Optimization by simulated annealing. cules. Genes Dev 15:1593 1612. Science 220:671–680. 45. Berthelot K, et al. (2004) Dynamics and processivity of 40S ribosome scanning on 31. Schuster P (1995) How to search for RNA structures. Theoretical concepts in evolution- mRNA in yeast. Mol Microbiol 51:987–1001. ary biotechnology. J Biotechnol 41:239–257. 46. Sambrook J, Fritsch EF, Maniatis T (1989) Molecular Cloning: A Laboratory Manual 32. Kertesz M, et al. (2010) Genome-wide measurement of RNA secondary structure in (Cold Spring Harbor Laboratory Press, New York). – yeast. Nature 467:103 107. 47. Stricker J, et al. (2008) A fast, robust and tunable synthetic gene oscillator. Nature 33. Washietl S, Hofacker IL, Stadler PF, Kellis M (2012) RNA folding with soft constraints: 456:516–519. Reconciliation of probing data and thermodynamic secondary structure prediction. 48. Takiff HE, Chen SM, Court DL (1989) Genetic analysis of the rnc operon of Escherichia Nucleic Acids Res 40:4261–4272. coli. J Bacteriol 171:2581–2590. 34. Rodrigo G, Carrera J, Jaramillo A (2011) Computational design of synthetic regulatory 49. Dunlop MJ, Cox RS, 3rd, Levine JH, Murray RM, Elowitz MB (2008) Regulatory activity networks from a genetic library to characterize the designability of dynamical beha- – viors. Nucleic Acids Res 39:e138. revealed by dynamic correlations in gene expression noise. Nat Genet 40:1493 1498. 35. Lease RA, Belfort M (2000) A trans-acting RNA as a control switch in Escherichia coli: 50. Paliy O, Gunasekera TS (2007) Growth of E. coli BL21 in minimal media with different DsrA modulates function by forming alternative structures. Proc Natl Acad Sci USA gluconeogenic carbon sources and salt contents. Appl Microbiol Biotechnol 97:9919–9924. 73:1169–1172.

15276 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1203831109 Rodrigo et al. Downloaded by guest on September 29, 2021